Title: AI-Generated 6G Internet Design: A Diffusion Model-based Learning Approach

Abstract: The 6G Internet poses intense demands for intelligent and customized designs to cope with the surging network scale, dynamically time-varying environments, diverse user requirements, and complicated manual configuration. However, traditional rule-based solutions heavily rely on human efforts and expertise, while data-driven intelligent algorithms still lack interpretability and generalization. In this paper, we propose the AIGI (AI-Generated Internet), a novel intention-driven design paradigm for the 6G Internet, which allows operators to quickly generate a variety of customized network solutions and achieve expert-free problem optimization. Driven by the diffusion model-based learning approach, AIGI has great potential to learn the reward-maximizing trajectories, automatically satisfy multiple constraints, adapt to different objectives and scenarios, or even intelligently create novel designs and mechanisms unseen in existing network environments. Finally, we conduct a use case to demonstrate that AIGI can effectively guide the design of transmit power allocation in digital twin-based 6G networks.